All-in-one (AIO) builds are a great way to perform an OpenStack-Ansible build for:
Although AIO builds aren’t recommended for large production deployments, they’re great for smaller proof-of-concept deployments.
Absolute minimum server resources (currently used for gate checks):
Recommended server resources:
bootstrap_host_data_disk_device
parameter. Please see
Building an AIO for more details.It’s possible to perform AIO builds within a virtual machine for demonstration and evaluation, but your virtual machines will perform poorly. For production workloads, multiple nodes for specific roles are recommended.
There are three steps to running an AIO build, with an optional first step should you need to customize your build:
When building an AIO on a new server, it is recommended that all system packages are upgraded and then reboot into the new kernel:
Note
Execute the following commands and scripts as the root user.
## Ubuntu
# apt-get dist-upgrade
# reboot
## CentOS
# yum upgrade
# yum install https://rdoproject.org/repos/openstack-ocata/rdo-release-ocata.rpm
# yum install git
# reboot
Note
If you are installing with limited connectivity, please review the Installing with limited connectivity appendix in the Deployment Guide before proceeding.
Start by cloning the OpenStack-Ansible repository and changing into the repository root directory:
# git clone https://git.openstack.org/openstack/openstack-ansible \
/opt/openstack-ansible
# cd /opt/openstack-ansible
Next switch the applicable branch/tag to be deployed from. Note that deploying from the head of a branch may result in an unstable build due to changes in flight and upstream OpenStack changes. For a test (for example, not a development) build it is usually best to checkout the latest tagged version.
# # List all existing tags. # git tag -l # # Checkout the stable branch and find just the latest tag # git checkout stable/ocata # git describe --abbrev=0 --tags # # Checkout the latest tag from either method of retrieving the tag. # git checkout 15.1.7
Note
The Ocata release is only compatible with Ubuntu 16.04 (Xenial Xerus) and Centos 7.
By default the scripts deploy all OpenStack services with sensible defaults for the purpose of a gate check, development or testing system.
Review the bootstrap-host role defaults file to see various configuration options. Deployers have the option to change how the host is bootstrapped. This is useful when you wish the AIO to make use of a secondary data disk, or when using this role to bootstrap a multi-node development environment.
The bootstrap script is pre-set to pass the environment variable
BOOTSTRAP_OPTS
as an additional option to the bootstrap process. For
example, if you wish to set the bootstrap to re-partition a specific
secondary storage device (/dev/sdb
), which will erase all of the data
on the device, then execute:
# export BOOTSTRAP_OPTS="bootstrap_host_data_disk_device=sdb"
Additional options may be implemented by simply concatenating them with a space between each set of options, for example:
# export BOOTSTRAP_OPTS="bootstrap_host_data_disk_device=sdb"
# export BOOTSTRAP_OPTS="${BOOTSTRAP_OPTS} bootstrap_host_ubuntu_repo=http://mymirror.example.com/ubuntu"
You may wish to change the role fetch mode. Options are galaxy
and
git-clone
. The default for this option is galaxy
.
galaxy: | Resolve all role dependencies using the ansible-galaxy resolver |
---|---|
git-clone: | Clone all of the role dependencies using native git |
ANSIBLE_ROLE_FETCH_MODE
to git-clone
. This will provide you the
ability to develop roles within the environment by modifying, patching, or
committing changes using an intact git tree while the galaxy
option
scrubs the .git
directory when it resolves a dependency.$ export ANSIBLE_ROLE_FETCH_MODE=git-clone
The next step is to bootstrap Ansible and the Ansible roles for the development environment. Deployers can customize roles by adding variables to override the defaults in each role (see Adding Galaxy roles). Run the following to bootstrap Ansible:
# scripts/bootstrap-ansible.sh
In order for all the services to run, the host must be prepared with the appropriate disks, packages, network configuration and a base configuration for the OpenStack Deployment. For the default AIO scenario, this preparation is completed by executing:
# scripts/bootstrap-aio.sh
If you wish to use a different scenario, for example, the Ceph scenario, execute the following:
# export SCENARIO='ceph'
# scripts/bootstrap-aio.sh
To add OpenStack Services over and above the bootstrap-aio default services
for the applicable scenario, copy the conf.d
files with the .aio
file
extension into /etc/openstack_deploy
and rename then to .yml
files.
For example, in order to enable the OpenStack Telemetry services, execute the
following:
cp etc/openstack_deploy/conf.d/{aodh,gnocchi,ceilometer}.yml.aio /etc/openstack_deploy/conf.d/
for f in $(ls -1 /etc/openstack_deploy/conf.d/*.aio); do mv -v ${f} ${f%.*}; done
To add any global overrides, over and above the defaults for the applicable
scenario, edit /etc/openstack_deploy/user_variables.yml
. See the
Deployment Guide for more details.
Finally, run the playbooks by executing:
# scripts/run-playbooks.sh
Note
Do not execute the run-playbooks.sh
more than once. If something goes
wrong, it is necessary to start over as described below in the
Rebuilding an AIO section. Alternatively, it may be possible to
individually run each playbook rather than starting over. If any playbooks
need to be re-run after the initial deploy, they should be run from the
playbooks directory with the openstack-ansible command. Executing
run-playbooks.sh
a second time results in an inconsistent state for LXC
IPtables rules and causes network connectivity issues from within containers.
The installation process will take a while to complete, but here are some general estimates:
Once the playbooks have fully executed, it is possible to experiment with
various settings changes in /etc/openstack_deploy/user_variables.yml
and
only run individual playbooks. For example, to run the playbook for the
Keystone service, execute:
# cd /opt/openstack-ansible/playbooks
# openstack-ansible os-keystone-install.yml
Note: The AIO bootstrap playbook will still build containers for services that are not requested for deployment, but the service will not be deployed in that container.
As the AIO includes all three cluster members of MariaDB/Galera, the cluster has to be re-initialized after the host is rebooted.
This is done by executing the following:
# cd /opt/openstack-ansible/playbooks
# openstack-ansible -e galera_ignore_cluster_state=true galera-install.yml
If this fails to get the database cluster back into a running state, then please make use of the Galera Cluster Recovery page in the Install Guide.
Sometimes it may be useful to destroy all the containers and rebuild the AIO. While it is preferred that the AIO is entirely destroyed and rebuilt, this isn’t always practical. As such the following may be executed instead:
# # Move to the playbooks directory.
# cd /opt/openstack-ansible/playbooks
# # Destroy all of the running containers.
# openstack-ansible lxc-containers-destroy.yml
# # On the host stop all of the services that run locally and not
# # within a container.
# for i in \
$(ls /etc/init \
| grep -e "nova\|swift\|neutron\|cinder" \
| awk -F'.' '{print $1}'); do \
service $i stop; \
done
# # Uninstall the core services that were installed.
# for i in $(pip freeze | grep -e "nova\|neutron\|keystone\|swift\|cinder"); do \
pip uninstall -y $i; done
# # Remove crusty directories.
# rm -rf /openstack /etc/{neutron,nova,swift,cinder} \
/var/log/{neutron,nova,swift,cinder}
# # Remove the pip configuration files on the host
# rm -rf /root/.pip
# # Remove the apt package manager proxy
# rm /etc/apt/apt.conf.d/00apt-cacher-proxy
Should an existing AIO environment need to be reinstalled, the most efficient method is to destroy the host operating system and start over. For this reason, AIOs are best run inside of some form of virtual machine or cloud guest.
Here is a basic diagram that attempts to illustrate what the resulting AIO deployment looks like.
This diagram is not to scale and is not even 100% accurate, this diagram was built for informational purposes only and should ONLY be used as such.
------->[ ETH0 == Public Network ]
|
V [ * ] Socket Connections
[ HOST MACHINE ] [ <>v^ ] Network Connections
* ^ *
| | |-------------------------------------------------------
| | |
| |---------------->[ HAProxy ] |
| ^ |
| | |
| V |
| (BR-Interfaces)<------- |
| ^ * | |
*-[ LXC ]*--*----------------------|-----|------|----| |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | V * | |
| * | | [ Galera x3 ] |
| [ Memcached ]<------------| | | |
*-------*[ Rsyslog ]<--------------|--| | * |
| [ Repos Server x3 ]<------| ---|-->[ RabbitMQ x3 ] |
| [ Horizon x2 ]<-----------| | | |
| [ Nova api ec2 ]<---------|--| | |
| [ Nova api os ]<----------|->| | |
| [ Nova console ]<---------| | | |
| [ Nova Cert ]<------------|->| | |
| [ Ceilometer api ]<-------|->| | |
| [ Ceilometer collector ]<-|->| | |
| [ Cinder api ]<-----------|->| | |
| [ Glance api ]<-----------|->| | |
| [ Heat apis ]<------------|->| | [ Loop back devices ]*-*
| [ Heat engine ]<----------|->| | \ \ |
| ------>[ Nova api metadata ] | | | { LVM } { XFS x3 } |
| | [ Nova conductor ]<-------| | | * * |
| |----->[ Nova scheduler ]--------|->| | | | |
| | [ Keystone x3 ]<----------|->| | | | |
| | |--->[ Neutron agents ]*-------|--|---------------------------*
| | | [ Neutron server ]<-------|->| | | |
| | | |->[ Swift proxy ]<----------- | | | |
*-|-|-|-*[ Cinder volume ]*----------------------* | |
| | | | | | |
| | | ----------------------------------------- | |
| | ----------------------------------------- | | |
| | -------------------------| | | | |
| | | | | | |
| | V | | * |
---->[ Compute ]*[ Neutron linuxbridge ]<---| |->[ Swift storage ]-
Except where otherwise noted, this document is licensed under Creative Commons Attribution 3.0 License. See all OpenStack Legal Documents.